Repeatability allows an estimation of the number of evaluations needed to optimize the selection of superior genotypes, with consequent effects on the research costs in terms of financial and human resources. The objective of this study was to estimate the coefficient of repeatability of biometric and yield traits, related to fruits and seeds of sacha inchi (Plukenetia volubilis), and to define the number of evaluations required for an efficient selection and evaluation of genotypes of the species. A total of 37 non-domesticated accessions were evaluated for 19 months in a randomized block design with 5 replications and 2 plants per plot. The total number of fruits, total number of seeds, total fruit weight, mean fruit weight, and number of seeds per fruit of the accessions were evaluated by monthly sampling. Additionally, seed biometry was assessed in a sample of 30 seeds per accession. Repeatability coefficients were estimated by analysis of variance, principal components and structural analysis. The principal component method based on the covariance matrix was the most appropriate for establishing repeatability estimates of sacha inchi, due to the cyclical nature of the crop. Superior genotypes of the species can be selected for yield-related traits with about 90% accuracy, from 5 harvests (months) onwards. To ensure this accuracy level, it would be necessary to evaluate a minimum of 5 and 25 fruits to determine mean fruit weight and number of seeds per fruit, respectively, and 39 seeds would be required to evaluate the biometric traits.